support for terra rasters in fasterize?

[jldupouey]

fasterize is a great package and I used it intensively with the couple sf/raster. But raster became terra. I checked for different cases, and fasterize is still very competitive in runtime with terra:rasterize (see an example below). But all of the fasterize time savings (and more) are lost in conversions between classes. Thus, it would be useful if fasterize could accept indifferently a vector of sf or SpatVector type, and a raster of raster or terra flavor as input. Is this conceivable?

MyVectTerr
#  class       : SpatVector 
#  geometry    : polygons 
#  dimensions  : 36570, 15  (geometries, attributes)
#  extent      : 99040, 1242445, 6046546, 7110480  (xmin, xmax, ymin, ymax)
#  ...

# rasterize with terra::rasterize:

MyRastTerr <- rast(MyVectTerr)
res(MyRastTerr) <- 100
system.time({
 Res1 <- rasterize(MyVectTerr, MyRastTerr, field="MyField")
 })
#    user  system elapsed 
#   13.81    0.97   14.80

# rasterize with fasterize:

MyVectSp <- as(MyVectTerr, "Spatial")  
MyVectSf <- st_as_sf(MyVectSp)

MyRastRaster <- raster(MyVectSf)
res(MyRastRaster) <- 100 
system.time({
 Res1 <- fasterize(MyVectSf, MyRastRaster, field="MyField")
 })
#    user  system elapsed 
#    1.90    0.04    1.95

[nikosGeography]

Any news regarding the issue?

[mdsumner]

it will take a fair bit of work to set up the writing out of blocks to a file. It's on my radar but with no timeline.

I'd prefer to fold the fasterize logic into GDAL itself, then it would be available to terra and other packages automatically also with memory management.

(I would recommend following up on what Even mentioned here: OSGeo/gdal#7200 (comment) that might help performance of terra, but final point - what about exactextractr? - this function should be pretty fast: https://isciences.gitlab.io/exactextractr/reference/rasterize_polygons.html)